arXiv subjects

Liyuan Liu

Publications and source records attributed to Liyuan Liu.

At least 37 records · Page 2Linked to original sources

arXiv2024

GRIN: GRadient-INformed MoE

Mixture-of-Experts (MoE) models scale more effectively than dense models due to sparse computation through expert routing, selectively activating only a small subset of expert modules. However, sparse computation challenges traditional training practices, as discrete expert routing hinders standard backpropagation and thus gradient-based optimization, which are the cornerstone of deep learning. To better pursue the scaling power of MoE, we introduce GRIN (GRadient-INformed MoE training), which incorporates sparse gradient estimation for expert routing and configures model parallelism to avoid token dropping. Applying GRIN to autoregressive language modeling, we develop a top-2 16$\times$3.8B MoE model. Our model, with only 6.6B activated parameters, outperforms a 7B dense model and matches the performance of a 14B dense model trained on the same data. Extensive evaluations across diverse tasks demonstrate the potential of GRIN to significantly enhance MoE efficacy, achieving 79.4 on MMLU, 83.7 on HellaSwag, 74.4 on HumanEval, and 58.9 on MATH.

cs.CL↗

arXiv2024

Model Tells Itself Where to Attend: Faithfulness Meets Automatic Attention Steering

Large language models (LLMs) have demonstrated remarkable performance across various real-world tasks. However, they often struggle to fully comprehend and effectively utilize their input contexts, resulting in responses that are unfaithful or hallucinated. This difficulty increases for contexts that are long or contain distracting information, which can divert LLMs from fully capturing essential evidence. To address this issue, many works use prompting to help LLMs utilize contextual information more faithfully. For instance, iterative prompting highlights key information in two steps that first ask the LLM to identify important pieces of context and then derive answers accordingly. However, prompting methods are constrained to highlighting key information implicitly in token space, which is often insufficient to fully steer the model's attention. To improve model faithfulness more reliably, we propose AutoPASTA, a method that automatically identifies key contextual information and explicitly highlights it by steering an LLM's attention scores. Like prompting, AutoPASTA is applied at inference time and does not require changing any model parameters. Our experiments on open-book QA demonstrate that AutoPASTA effectively enables models to grasp essential contextual information, leading to substantially improved model faithfulness and performance, e.g., an average improvement of 7.95% for LLAMA3-70B-Instruct. Code will be publicly available at https://github.com/QingruZhang/AutoPASTA .

cs.CL↗

arXiv2024

Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone

We introduce phi-3-mini, a 3.8 billion parameter language model trained on 3.3 trillion tokens, whose overall performance, as measured by both academic benchmarks and internal testing, rivals that of models such as Mixtral 8x7B and GPT-3.5 (e.g., phi-3-mini achieves 69% on MMLU and 8.38 on MT-bench), despite being small enough to be deployed on a phone. Our training dataset is a scaled-up version of the one used for phi-2, composed of heavily filtered publicly available web data and synthetic data. The model is also further aligned for robustness, safety, and chat format. We also provide parameter-scaling results with a 7B, 14B models trained for 4.8T tokens, called phi-3-small, phi-3-medium, both significantly more capable than phi-3-mini (e.g., respectively 75%, 78% on MMLU, and 8.7, 8.9 on MT-bench). To enhance multilingual, multimodal, and long-context capabilities, we introduce three models in the phi-3.5 series: phi-3.5-mini, phi-3.5-MoE, and phi-3.5-Vision. The phi-3.5-MoE, a 16 x 3.8B MoE model with 6.6 billion active parameters, achieves superior performance in language reasoning, math, and code tasks compared to other open-source models of similar scale, such as Llama 3.1 and the Mixtral series, and on par with Gemini-1.5-Flash and GPT-4o-mini. Meanwhile, phi-3.5-Vision, a 4.2 billion parameter model derived from phi-3.5-mini, excels in reasoning tasks and is adept at handling both single-image and text prompts, as well as multi-image and text prompts.

cs.CL↗

arXiv2024

Learning a Decision Tree Algorithm with Transformers

Decision trees are renowned for their ability to achieve high predictive performance while remaining interpretable, especially on tabular data. Traditionally, they are constructed through recursive algorithms, where they partition the data at every node in a tree. However, identifying a good partition is challenging, as decision trees optimized for local segments may not yield global generalization. To address this, we introduce MetaTree, a transformer-based model trained via meta-learning to directly produce strong decision trees. Specifically, we fit both greedy decision trees and globally optimized decision trees on a large number of datasets, and train MetaTree to produce only the trees that achieve strong generalization performance. This training enables MetaTree to emulate these algorithms and intelligently adapt its strategy according to the context, thereby achieving superior generalization performance.

cs.LG↗

arXiv2024

Toward Student-Oriented Teacher Network Training For Knowledge Distillation

How to conduct teacher training for knowledge distillation is still an open problem. It has been widely observed that a best-performing teacher does not necessarily yield the best-performing student, suggesting a fundamental discrepancy between the current teacher training practice and the ideal teacher training strategy. To fill this gap, we explore the feasibility of training a teacher that is oriented toward student performance with empirical risk minimization (ERM). Our analyses are inspired by the recent findings that the effectiveness of knowledge distillation hinges on the teacher's capability to approximate the true label distribution of training inputs. We theoretically establish that the ERM minimizer can approximate the true label distribution of training data as long as the feature extractor of the learner network is Lipschitz continuous and is robust to feature transformations. In light of our theory, we propose a teacher training method SoTeacher which incorporates Lipschitz regularization and consistency regularization into ERM. Experiments on benchmark datasets using various knowledge distillation algorithms and teacher-student pairs confirm that SoTeacher can improve student accuracy consistently.

cs.LG↗

arXiv2024

DHRNet: A Dual-Path Hierarchical Relation Network for Multi-Person Pose Estimation

Multi-person pose estimation (MPPE) presents a formidable yet crucial challenge in computer vision. Most existing methods predominantly concentrate on isolated interaction either between instances or joints, which is inadequate for scenarios demanding concurrent localization of both instances and joints. This paper introduces a novel CNN-based single-stage method, named Dual-path Hierarchical Relation Network (DHRNet), to extract instance-to-joint and joint-to-instance interactions concurrently. Specifically, we design a dual-path interaction modeling module (DIM) that strategically organizes cross-instance and cross-joint interaction modeling modules in two complementary orders, enriching interaction information by integrating merits from different correlation modeling branches. Notably, DHRNet excels in joint localization by leveraging information from other instances and joints. Extensive evaluations on challenging datasets, including COCO, CrowdPose, and OCHuman datasets, showcase DHRNet's state-of-the-art performance. The code will be released at https://github.com/YHDang/dhrnet-multi-pose-estimation.

cs.CV↗

arXiv2023

Bridging Discrete and Backpropagation: Straight-Through and Beyond

Backpropagation, the cornerstone of deep learning, is limited to computing gradients for continuous variables. This limitation poses challenges for problems involving discrete latent variables. To address this issue, we propose a novel approach to approximate the gradient of parameters involved in generating discrete latent variables. First, we examine the widely used Straight-Through (ST) heuristic and demonstrate that it works as a first-order approximation of the gradient. Guided by our findings, we propose ReinMax, which achieves second-order accuracy by integrating Heun's method, a second-order numerical method for solving ODEs. ReinMax does not require Hessian or other second-order derivatives, thus having negligible computation overheads. Extensive experimental results on various tasks demonstrate the superiority of ReinMax over the state of the art. Implementations are released at https://github.com/microsoft/ReinMax.

cs.LG↗

arXiv2023

Label Noise in Adversarial Training: A Novel Perspective to Study Robust Overfitting

We show that label noise exists in adversarial training. Such label noise is due to the mismatch between the true label distribution of adversarial examples and the label inherited from clean examples - the true label distribution is distorted by the adversarial perturbation, but is neglected by the common practice that inherits labels from clean examples. Recognizing label noise sheds insights on the prevalence of robust overfitting in adversarial training, and explains its intriguing dependence on perturbation radius and data quality. Also, our label noise perspective aligns well with our observations of the epoch-wise double descent in adversarial training. Guided by our analyses, we proposed a method to automatically calibrate the label to address the label noise and robust overfitting. Our method achieves consistent performance improvements across various models and datasets without introducing new hyper-parameters or additional tuning.

cs.LG↗

arXiv2023

Fast-ELECTRA for Efficient Pre-training

ELECTRA pre-trains language models by detecting tokens in a sequence that have been replaced by an auxiliary model. Although ELECTRA offers a significant boost in efficiency, its potential is constrained by the training cost brought by the auxiliary model. Notably, this model, which is jointly trained with the main model, only serves to assist the training of the main model and is discarded post-training. This results in a substantial amount of training cost being expended in vain. To mitigate this issue, we propose Fast-ELECTRA, which leverages an existing language model as the auxiliary model. To construct a learning curriculum for the main model, we smooth its output distribution via temperature scaling following a descending schedule. Our approach rivals the performance of state-of-the-art ELECTRA-style pre-training methods, while significantly eliminating the computation and memory cost brought by the joint training of the auxiliary model. Our method also reduces the sensitivity to hyper-parameters and enhances the pre-training stability.

cs.CL↗

arXiv2023

Understanding the Difficulty of Training Transformers

Transformers have proved effective in many NLP tasks. However, their training requires non-trivial efforts regarding designing cutting-edge optimizers and learning rate schedulers carefully (e.g., conventional SGD fails to train Transformers effectively). Our objective here is to understand $\textit{what complicates Transformer training}$ from both empirical and theoretical perspectives. Our analysis reveals that unbalanced gradients are not the root cause of the instability of training. Instead, we identify an amplification effect that influences training substantially -- for each layer in a multi-layer Transformer model, heavy dependency on its residual branch makes training unstable, since it amplifies small parameter perturbations (e.g., parameter updates) and results in significant disturbances in the model output. Yet we observe that a light dependency limits the model potential and leads to inferior trained models. Inspired by our analysis, we propose Admin ($\textbf{Ad}$aptive $\textbf{m}$odel $\textbf{in}$itialization) to stabilize stabilize the early stage's training and unleash its full potential in the late stage. Extensive experiments show that Admin is more stable, converges faster, and leads to better performance. Implementations are released at: https://github.com/LiyuanLucasLiu/Transforemr-Clinic.

cs.LG↗

arXiv2023

Sparse Backpropagation for MoE Training

One defining characteristic of Mixture-of-Expert (MoE) models is their capacity for conducting sparse computation via expert routing, leading to remarkable scalability. However, backpropagation, the cornerstone of deep learning, requires dense computation, thereby posting challenges in MoE gradient computations. Here, we introduce SparseMixer, a scalable gradient estimator that bridges the gap between backpropagation and sparse expert routing. Unlike typical MoE training which strategically neglects certain gradient terms for the sake of sparse computation and scalability, SparseMixer provides scalable gradient approximations for these terms, enabling reliable gradient estimation in MoE training. Grounded in a numerical ODE framework, SparseMixer harnesses the mid-point method, a second-order ODE solver, to deliver precise gradient approximations with negligible computational overhead. Applying SparseMixer to Switch Transformer on both pre-training and machine translation tasks, SparseMixer showcases considerable performance gain, accelerating training convergence up to 2 times.

cs.LG↗

arXiv2023

Parallel Data Compression Techniques

With endless amounts of data and very limited bandwidth, fast data compression is one solution for the growing datasharing problem. Compression helps lower transfer times and save memory, but if the compression takes too long, this no longer seems viable. Multi-core processors enable parallel data compression; however, parallelizing the algorithms is anything but straightforward since compression is inherently serial. This paper explores techniques to parallelize three compression schemes: Huffman coding, LZSS, and MP3 coding

cs.DC↗

arXiv2022

P4E: Few-Shot Event Detection as Prompt-Guided Identification and Localization

We propose P4E, an identify-and-localize event detection framework that integrates the best of few-shot prompting and structured prediction. Our framework decomposes event detection into an identification task and a localization task. For the identification task, which we formulate as multi-label classification, we leverage cloze-based prompting to align our objective with the pre-training task of language models, allowing our model to quickly adapt to new event types. We then employ an event type-agnostic sequence labeling model to localize the event trigger conditioned on the identification output. This heterogeneous model design allows P4E to quickly learn new event types without sacrificing the ability to make structured predictions. Our experiments demonstrate the effectiveness of our proposed design, and P4E shows superior performance for few-shot event detection on benchmark datasets FewEvent and MAVEN and comparable performance to SOTA for fully-supervised event detection on ACE.

cs.CL↗

arXiv2021

On the Variance of the Adaptive Learning Rate and Beyond

The learning rate warmup heuristic achieves remarkable success in stabilizing training, accelerating convergence and improving generalization for adaptive stochastic optimization algorithms like RMSprop and Adam. Here, we study its mechanism in details. Pursuing the theory behind warmup, we identify a problem of the adaptive learning rate (i.e., it has problematically large variance in the early stage), suggest warmup works as a variance reduction technique, and provide both empirical and theoretical evidence to verify our hypothesis. We further propose RAdam, a new variant of Adam, by introducing a term to rectify the variance of the adaptive learning rate. Extensive experimental results on image classification, language modeling, and neural machine translation verify our intuition and demonstrate the effectiveness and robustness of our proposed method. All implementations are available at: https://github.com/LiyuanLucasLiu/RAdam.

cs.LG↗

arXiv2021

Data Quality Matters For Adversarial Training: An Empirical Study

Multiple intriguing problems are hovering in adversarial training, including robust overfitting, robustness overestimation, and robustness-accuracy trade-off. These problems pose great challenges to both reliable evaluation and practical deployment. Here, we empirically show that these problems share one common cause -- low-quality samples in the dataset. Specifically, we first propose a strategy to measure the data quality based on the learning behaviors of the data during adversarial training and find that low-quality data may not be useful and even detrimental to the adversarial robustness. We then design controlled experiments to investigate the interconnections between data quality and problems in adversarial training. We find that when low-quality data is removed, robust overfitting and robustness overestimation can be largely alleviated; and robustness-accuracy trade-off becomes less significant. These observations not only verify our intuition about data quality but may also open new opportunities to advance adversarial training.

cs.LG↗

arXiv2021

On the Transformer Growth for Progressive BERT Training

Due to the excessive cost of large-scale language model pre-training, considerable efforts have been made to train BERT progressively -- start from an inferior but low-cost model and gradually grow the model to increase the computational complexity. Our objective is to advance the understanding of Transformer growth and discover principles that guide progressive training. First, we find that similar to network architecture search, Transformer growth also favors compound scaling. Specifically, while existing methods only conduct network growth in a single dimension, we observe that it is beneficial to use compound growth operators and balance multiple dimensions (e.g., depth, width, and input length of the model). Moreover, we explore alternative growth operators in each dimension via controlled comparison to give operator selection practical guidance. In light of our analyses, the proposed method speeds up BERT pre-training by 73.6% and 82.2% for the base and large models respectively, while achieving comparable performances

cs.CL↗

arXiv2021

Empower Distantly Supervised Relation Extraction with Collaborative Adversarial Training

With recent advances in distantly supervised (DS) relation extraction (RE), considerable attention is attracted to leverage multi-instance learning (MIL) to distill high-quality supervision from the noisy DS. Here, we go beyond label noise and identify the key bottleneck of DS-MIL to be its low data utilization: as high-quality supervision being refined by MIL, MIL abandons a large amount of training instances, which leads to a low data utilization and hinders model training from having abundant supervision. In this paper, we propose collaborative adversarial training to improve the data utilization, which coordinates virtual adversarial training (VAT) and adversarial training (AT) at different levels. Specifically, since VAT is label-free, we employ the instance-level VAT to recycle instances abandoned by MIL. Besides, we deploy AT at the bag-level to unleash the full potential of the high-quality supervision got by MIL. Our proposed method brings consistent improvements (~ 5 absolute AUC score) to the previous state of the art, which verifies the importance of the data utilization issue and the effectiveness of our method.

cs.CL↗

arXiv2021

Multi-head or Single-head? An Empirical Comparison for Transformer Training

Multi-head attention plays a crucial role in the recent success of Transformer models, which leads to consistent performance improvements over conventional attention in various applications. The popular belief is that this effectiveness stems from the ability of jointly attending multiple positions. In this paper, we first demonstrate that jointly attending multiple positions is not a unique feature of multi-head attention, as multi-layer single-head attention also attends multiple positions and is more effective. Then, we suggest the main advantage of the multi-head attention is the training stability, since it has less number of layers than the single-head attention, when attending the same number of positions. For example, 24-layer 16-head Transformer (BERT-large) and 384-layer single-head Transformer has the same total attention head number and roughly the same model size, while the multi-head one is significantly shallower. Meanwhile, we show that, with recent advances in deep learning, we can successfully stabilize the training of the 384-layer Transformer. As the training difficulty is no longer a bottleneck, substantially deeper single-head Transformer achieves consistent performance improvements without tuning hyper-parameters.

cs.CL↗